本文提出了一种机器人视觉系统中有效识别三维物体的方法。利用狭缝光束离散投影的少量样本数据进行特征提取，实现高速扫描。通过基于边界线搜索模型以匹配描述来识别对象。实验结果表明，该方法可以提取物体角点的顶点位置和角度，精度为1MM和2度。此外，可以使用几种不同模型中的一种来正确识别对象。

THIS PAPER PRESENTS A METHOD TO RECOGNIZE 3-D OBJECTS EFFECTIVELY FOR A ROBOT VISION SYSTEM. THE FEATURE EXTRACTION CAN BE CARRIED OUT BY LESS SAMPLE DATA FROM DISCRETE PROJECTION OF SLIT BEAMS FOR HIGH-SPEED SCANNING. THE OBJECT IS IDENTIFIED BY SEARCHING A MODEL TO MATCH THE DESCRIPTION BASED ON THE BOUNDARY LINES. AS THE RESULT OF THE EXPERIMENT, THE VERTEX POSITIONS AND ANGLES AT THE CORNERS OF THE OBJECT CAN BE EXTRACTED WITH AN ACCURACY OF 1MM AND 2 DEGREES. ALSO, THE OBJECT CAN BE CORRECTLY IDENTIFIED WITH ONE OF THE SEVERAL DIFFERENT MODELS.